The codes developed and distributed in the context of the NLAFET project are hosted on the GitHub platform and are available via the link GitHub NLAFET. In the table below, the public repositories that together constitute the NLAFET library software are listed and organized according to the following topics:
- Dense matrix factorizations and solvers
- Solvers and tools for standard and generalized dense eigenvalue problems
- Sparse direct factorizations and solvers
- Communication optimal algorithms for iterative methods
- Cross-cutting tools
|Repository name||Brief description|
|BBLAS-ref||Batched BLAS reference implementation|
|plasma||Snapshot of PLASMA library|
|PlaStar||Some StarPU implementations|
|SEVP-PDHSEQR-Alg953||Standard eigenvalue problem solvers|
|GEVP-PDHGEQZ||Generalized eigenvalue problem solvers|
|StarNEig||StarPU nonsymmetric eigenvalue problem solvers|
|SpLLT||Sparse LLT solver for A = AT, positive definite|
|SyLVER||Symmetrically stuctured factorizations|
|ParSHUM||Sparse LU solver for highly unsymmetric matrices|
|BC||Hybrid solver based on block Cimmino for square and rectangular systems|
|preAlps||Preconditioned iterative methods and enlarged Krylov method|
|PCP-runtime||Parallelizing the critical path|
|ABFT-factor||Tiled factorizations with ABFT|
- The parallel programming models used include MPI, OpenMP, StarPU, and ParSEC.
- By using task-based runtime systems we aim at strongly parallel software for heterogeneous extreme scale systems.
- For more information see the D7.8 deliverable Release of the NLAFET library.